Editor's note: At the University of California, San Francisco, Robert M. Wachter, MD, is Professor and Chief of the Division of Hospital Medicine; Associate Chairman of the Department of Medicine; Lynne and Marc Benioff Endowed Chair in Hospital Medicine; and Chief of the Medical Service at UCSF Medical Center. He is also Editor of AHRQ WebM&M and AHRQ Patient Safety Network. A leading expert on hospital medicine as well as patient safety, he was lead author on a recent commentary in the Joint Commission Journal on Quality and Patient Safety and coauthored a recent commentary in the Journal of the American Medical Association (JAMA) on insurers not paying for preventable complications. We asked him to speak with us about the benefits and limitations of such changes.

Dr. Niraj Sehgal, Associate Editor, AHRQ WebM&M: You've written a couple of recent articles about Medicare's decision to withhold payment for preventable adverse events. While you've applauded their efforts to drive improvement, you've also raised caution about the details of the policy. What concerns you most about their initiative?

Dr. Robert Wachter: The idea makes a lot of sense. It's an artifact of Medicare's DRG [diagnosis-related group] payment system that adverse events lead to higher payments. That was appropriate when the adverse events just happened and they cost institutions more money to take care of sicker patients. But we've come to recognize that some adverse events are at least partly preventable if we do the right things all the time, and so a policy of not paying extra for them makes sense. My concern is that, for the policy to be appropriate and just, the events really do need to be preventable. We also need to be able to figure out whether patients have the adverse event when they come in the door of a hospital or whether it developed on the hospital's watch. We need to be able to figure out a way of measuring the events accurately so that the tally of adverse events isn't gameable. At this point, the science is lagging behind the policy. As you look at the initial list of adverse events on the Medicare list, the number of events that meet all of those conditions is really very small.

NS: You described several criteria that would allow a condition to be included on this list of preventable adverse events. Talk a little bit about the concept of "present on admission," something you've written about, and how important that specific criterion is as to whether an adverse event should or should not be included on the list.

RW: Present at admission is a very interesting construct that is pretty new. If you are going to promulgate a policy that either publicly reports adverse events that happen in hospitals or pay differentially when you have them, of course you have to be sure that the hospital is not being blamed, for lack of a better word, for things that didn't happen on their watch. That's particularly important for big academic medical centers. My hospital, UCSF Medical Center, gets 10% to 15% of its admissions from other hospitals. Sometimes patients are sent to us because they've had bad things happen—they're very ill and it's our job to provide a higher level of care or more specialized care for those disorders. So if the hospital is going to be tagged with an adverse event that happened at another place, that's profoundly unfair. A recent study looked at the AHRQ Patient Safety Indicators and found that many patients already had these adverse events when they arrived. Not accounting for "present on admission" unfairly made the hospitals look bad on the Patient Safety Indicators. For me, an apt analogy is that before I drive off the lot at Hertz, the guy walks around my car and looks all over the car to be sure there were no dings when I left the lot. So we're going to have to have the equivalent of the Hertz guy walking around the patients looking at their bumpers when they're in our emergency department or our admitting office, to see whether they already have those dents before they come in the hospital. [See related commentary.]

NS: Can you comment on the tension between documentation of care and actual care delivered? You can envision that many hospitals may figure out ways to document care well, but may in fact not be actually delivering that care.

RW: Well, we certainly have seen that. Probably the best example is one of the quality measures that has been promoted over the last few years: smoking cessation counseling. Hospitals have done better on that measure. But the measure is captured by looking to see if there is a note in the chart that says that smoking cessation counseling happened. If you look at 100 patients' charts that have documentation of smoking cessation counseling, I would be shocked if more than 10% or 20% of those patients truly received what we consider really effective smoking cessation counseling. I'm not accusing the hospitals or the doctors of lying. I think that probably in most cases someone does say to the patient, "Do you smoke?" And the patient says, "Yes, I do," and the doctor says, "You shouldn't do that." At the end of that conversation, someone will check a box and say that smoking cessation counseling happened—as they do about advance directives and other elements of complex care that take time, where the metric of whether it happened is whether a box was checked. When something that might prevent an adverse event can be documented through a check box, hospitals will develop systems to push the check box in front of the providers and make it very hard to move to the next stage of clinical care if you have not, in fact, checked the box. At some level, it gets the issue front and center in the eyes of providers. But to the extent that checking the box is supposed to be a proxy for something useful happening in patient care, we have to recognize that a lot of the time it's a poor representation of what actually happens, and it's something that can be and I think is often gamed.

NS: Many patient safety advocate groups have claimed that the new Medicare policy of "not paying for errors" is long overdue. That in fact, hospitals should be paying for things that shouldn't happen, and that this type of policy, while perhaps overstretching, really will drive change. Do you see evidence of this policy actually creating some change at the system level?

RW: I agree with those people. I actually think the policy is a good idea. If you take the old situation and look at it, I think you'll realize that it's a little bit odd. Let's think about central line–associated bloodstream infections, which we know with good scientific evidence can be prevented if hospitals reliably perform an evidence-based set of practices, none of which is all that onerous. We can prevent this really bad thing from happening, which, when it happens, is harmful and often costs lives. Under the old payment system, when the adverse event happened, the hospital would get paid an extra 50% or so by Medicare. Now, some people have taken offense at this whole line of reasoning and say, "Does Medicare think that I, as a doctor, or I, as a hospital CEO, was trying to harm patients?" Of course not. What they do believe, and I share this belief, is that we have come to recognize that patient safety is often a systems manifestation. Improving systems costs money, time, focus, and political capital, and that money is not likely to be spent, the focus is not likely to occur, if there is no business case for change. So the idea of a policy that creates skin in the game for hospitals to do the right thing and promotes these kinds of preventive activities is completely sensible and completely ethical.

The problem is that the devil is in the details. These new policies always lead to unforeseen consequences, and we cannot be so risk averse that we don't act until we understand this perfectly. What I do believe, though, is that if Medicare goes too far or too fast, or if other insurers jump on with policies that add additional diagnoses or don't account for present on admission, or the definitions of the adverse events are too variable, the risk is that providers, both institutions and individual providers, will look at this policy and wonder whether this is really about patient safety and preventing medical errors, or whether it is just a cost-cutting maneuver being clothed in the very nice garb of patient safety. I don't think that's what Medicare intends, but I think that's the risk of going too far and too fast with this.

NS: Do you think the Medicare policy is really about saving lives and improving care, or simply cutting costs?

RW: I think a little bit of both. If you were Medicare and you were looking at the impending bankruptcy of the Medicare trust fund, it would defy the laws of administrative gravity for people in Medicare not to be constantly looking at their policies through a lens of saving money. They have to be thinking that. My guess is that the answer is and, rather than or—that as they look at their payment policy and they realize that they are paying hospitals more when something bad happens that could have been prevented, they see that as an opportunity for a "twofer." That not only can we improve safety and decrease the rate of these adverse events, but at the same time, why should we be using scarce tax dollars paying for these things? The problem is that the conflation of these two goals leads to the possibility that somebody's going to forget about goal one, which is patient safety and preventing adverse events, in the name of goal two.

NS: If you were going to look in your crystal ball, say 5 or 10 years ahead, how do you see all this playing out with the Medicare policy as it grows into effect? Do you think we're going to see more conditions added, some of them removed, or the whole thing implode in some capacity?

RW: The overall policy direction is so sensible that I cannot see it going away. Part of that is predicated on whether the payment system looks like today's payment system. But if, in fact, the Medicare payment system remains a DRG-like system where we will pay for a given diagnosis, I think it is inevitable that when adverse events happen, which are completely or largely preventable, that they won't be paid for. I think the science will be substantially better than it is now. We will much more fully understand how to prevent some of these events. It's likely that in addition to a "no pay for errors" policy, there will probably also be some element of pay for performance, which there really hasn't been in patient safety (as opposed to quality).

NS: Let me ask for a bit of practical guidance. Let's say I was the chief of a medical service or the leader of a hospital medicine group, both roles you currently have. How do you plan to prepare your service and hospitalists or other physicians for the implications of these policies? Is there something in particular you have to be doing differently as this policy goes into effect?

RW: That's a great question. It would be important for every physician for whom this is relevant to understand the nature of the policy, why it's in place, and which of these diagnoses truly are preventable and what they can do to help prevent them. They should also understand the construct of "present on admission." That said, one of the things we've learned in the patient safety and quality revolution over the past 10 years is that trying to educate each of the 1000 physicians or 5000 nurses to each do the right thing every time is probably the weakest method of systems change. Most of our successes in both the safety and quality fields have not come from the old notion of continuing medical education or nursing education and trying to get everybody to do the same thing every time by teaching them about it. Rather, the successes have come from systematizing practice. For example, asking the individual physicians to remember the five or ten or fifteen diagnoses and figure out whether they were present on admission is probably futile, if that's not backed up by a system that essentially forces them to do that. Whether the system will be a series of hard stops that are built into electronic health records or involves a newly hired, present-on-admission nurse, whose job is to be the Hertz guy and look the patient over to determine whether the patient has any of these ten things on admission, I don't know. But I think that there will be some element of systemization of these kinds of things. What we've learned from Pronovost's work in Michigan and other studies in preventing central line–associated bloodstream infections is that the systemization of preventive activities is where most of the mileage will be. Trying to get every doctor or nurse to remember to do these five things to prevent catheter-associated bloodstream infections, these seven things to prevent falls, and these nine things to prevent decubitus ulcers—that's too much for any human brain to handle. Rather than trying to ask them to remember everything, ask them what does the system look like that would guarantee that we would get this stuff?

NS: I walk around the hospital these days and it seems that people are increasingly saying, "We need to address this as a safety issue." Do you think this is a good thing? Do you worry that everything today is being described as a safety issue because it's kind of in vogue?

RW: It's both. The first step in creating the patient safety revolution was to get patient safety on the radar screen and to have it be part of the clinical and moral vocabulary of medicine. And I think that's happened. Many of the things that people talk about as safety issues in fact are. Now that doesn't mean that we can achieve them all. When I was an intern, if I was really tired and had a nasty attitude at the end of being up 36 hours in a row, I would have labeled that as, I'm really tired and cranky. Now that is defined by our house staff and others as a safety issue. Well, in fact, it is. What we don't really know is how to fix that. The resources that we have to play with are relatively fixed. As we shrink the house staff hours to try to make sure that they're less tired—so hopefully patients will be safer and the house staff will be less cranky—we've seen increasing handoffs and the beginning of a shift mentality. These are unforeseen consequences that pop out the back end. It's sort of nice and easy and morally pure to say that once you've defined this as a safety issue, how can you not spend whatever it takes to fix that. But then you have that old "no money, no mission" problem that they talk about in the church. If you fix all of the issues that are defined as safety issues, you cannot, at least in the modern payment system, run a hospital profitably. So defining the safety issues is fine and reasonable. It raises the visibility. But we're not very good yet at taking all of these "safety issues," processing them, and prioritizing which ones we have to fix today, which ones we should fix tomorrow, and which ones we define as yes, things would be a little bit safer if we did it a different way. We're just beginning to learn how to do this kind of prioritization.

NS: Let me take you back to your crystal ball, particularly to your political lens, and it's now February 2009 and your phone rings and the new president asks you to write a policy that helps drive improvement and ensures that truly preventable adverse events don't occur in hospitals. One, do you take the phone call, and two, what would you write as a policy?

RW: Yes, I would take the phone call. To me the fundamental reason that care is not nearly as safe as it should be is that up until 7 or 8 years ago, there was precisely no incentive for medical schools, nursing schools, individual doctors, individual nurses, pharmacists, or hospitals to focus on safety, other than professionalism. That's too weak an incentive to generate the kinds of changes that we want. What I would tell them is the lesion right now is the lack of evidence. I don't believe that we need many more policies promoting public reporting and transparency, or pay for performance or no pay for errors. The policy pipeline has been pretty thoughtful about what things we should do. Where we're outrunning the blood supply is that these policies only work when we can, in a generally agreed upon and in an auditable way, call something ventilator-associated pneumonia and know that we're all talking about the same thing. When hospitals can reach into a tool kit and with good evidence know they have to do these five things, and if they did them religiously, that would prevent 90% of these bad things from happening. It's not effective to have 6000 hospitals all trying different things with no mechanism to learn from their experiences. The real lesion in our approach is that we spend 30 billion dollars on scientific research at the NIH (and if you add in the pharmaceutical companies' and device companies' investment, it's probably two or three times that) all trying to advance the science. We spend a fraction of that—one fiftieth or one hundredth of that when you look at the budget of AHRQ and other organizations focused on implementation science—focused on, do we know with good scientific evidence that if you do these five things you will prevent most decubitus ulcers, or most falls. That's where we need to be investing if we're going to promote patient safety. The policies are just scaffolding on which you hang the evidence-based science for doing the right thing. Until we have that science, the policies can only do so much. And the chances that the policies are going to cause mischief and unintended consequences, I think, are unacceptably high.